The Quality Spreadsheet Stops Living on Its Own

The problem

There are two truths about quality, and the good one is not in the ERP.

The real spreadsheet, holding live data on weight, filling ratio, batter moisture and core temperature, and the official ERP, updated weeks later by someone spending hours a month copying and pasting.

• The real spreadsheet carries the live numbers: piece weight, filling ratio, batter moisture and core temperature after the tunnel, run by run.
• The official ERP carries whatever somebody copied into it weeks later, after spending hours a month on the copying itself.
• For a gourmet range that is not a formality: the filling ratio is the claim printed on the pack, the pack carries your own brand, and a customer who bought it directly from your store is the one comparing the two.
• Meanwhile several lab staff hold an ERP seat purely to key in numbers that already exist, correctly, in a file that nobody outside the laboratory can query.

How it fits the IRIS system

Connect on the shared folder — it reads the document, not the person.

iLEAN watches the shared folder and reads the delta the moment the file is saved, parsing each cell with an anchored LLM. It reads the document, not the person. If several lab staff log into the ERP today just to enter quality data, with Connect only the validator needs a seat. Typical reduction of 50 to 80% in quality-module licenses.

The lab keeps working in the same spreadsheet, each analyst in their own way. iLEAN reads the delta the moment the file is saved and parses each cell with an anchored model. What disappears is the transcription, and with it the weeks of drift between the sheet the plant trusts and the system the business reports from.

See the full IRIS architecture →

Before and after

Today's quality data versus quality data connected

Aspect	Today	With iLEAN Connect
Gap between sheet and ERP	Weeks	Second zero
Filling ratio per run	In a file nobody queries	Queryable against the batch
Core temperature after the tunnel	Typed twice, sometimes	Read once, pushed once
Who keys it in	Several lab staff	Nobody: only a validator
Quality module licenses	One per person entering data	50-80% fewer
The spreadsheet's format	—	Unchanged

Real spreadsheet and official ERP weeks out of sync → Quality data in the ERP in zero seconds, a single license.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 4-9 months, including the recurring annual saving on ERP licenses, and to be validated against the real invoice rather than a list price.
• A typical reduction of 50 to 80% in quality-module seats, because only the validator needs one and everybody else keeps working in the file they already use.
• Hours a month of copy-and-paste returned to the laboratory, and a quality history that stops being weeks out of date.
• And the filling ratio behind a pack claim becomes queryable per run, instead of being rebuilt from a file whenever a distributor or an auditor asks for it.

Payback 4-9 months · -50/-80% quality-module ERP licenses. Estimated payback is 4 to 9 months, including the recurring annual savings on ERP licenses. Estimate to validate against the real ERP invoice.

And the fair question from the production manager

"Our spreadsheet is a mess: no two analysts fill it in the same way, and half of it is comments in the margin." — that is the normal case, and the reason this works: reading known fields out of a document is an anchored task where the best models drop below 1.5% error [1], and the meaning of a field matters far more than which cell it happens to sit in. There is also a second net: out-of-specification values are never pushed silently, they are raised for the quality lead before anything reaches the ERP.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.